Skip to main content

Foundational Trials

Terrangi

Image: Shutterstock / ChaNaWiT

Publication

  • Title: Early vs Late Tracheotomy for Prevention of Pneumonia in Mechanically Ventilated Adult ICU Patients: A Randomized Controlled Trial
  • Acronym: Not applicable
  • Year: 2010
  • Journal published in: JAMA
  • Citation: Terragni PP, Antonelli M, Fumagalli R, Faggiano C, Berardino M, Pallavicini FB, et al. Early vs late tracheotomy for prevention of pneumonia in mechanically ventilated adult ICU patients: a randomized controlled trial. JAMA. 2010;303(15):1483-1489.

Context & Rationale

  • Background
    • Tracheotomy is frequently used to replace translaryngeal intubation when prolonged mechanical ventilation is anticipated, aiming to improve comfort, reduce sedative exposure, facilitate weaning, and optimise secretion clearance.
    • Ventilator-associated pneumonia (VAP) is a clinically important complication of prolonged ventilation; tracheotomy timing was biologically plausible as a modifier of VAP risk (e.g., airway colonisation, microaspiration, secretion management).
    • Before this trial, evidence for “early” tracheotomy was limited by heterogeneity in definitions (timing thresholds), patient selection, and outcome ascertainment, with inconsistent signals for VAP and mortality.
    • Because tracheotomy carries procedural risks, a credible demonstration of benefit (particularly for infection prevention or hard outcomes) was needed to justify routinely accelerating the procedure.
  • Research Question/Hypothesis
    • In mechanically ventilated adult ICU patients with persistent severe respiratory dysfunction after 48 hours and without pneumonia, does early tracheotomy (day 6–8 after intubation) reduce the 28-day cumulative incidence of VAP compared with late tracheotomy (day 13–15)?
    • Secondary hypothesis: earlier tracheotomy would improve process outcomes (earlier ventilator liberation and ICU discharge) without increasing harms.
  • Why This Matters
    • Timing of tracheotomy is a high-frequency ICU decision with major implications for sedation strategy, weaning trajectories, nursing workload, and resource use.
    • If earlier timing meaningfully reduced VAP or accelerated liberation, it could change practice; conversely, null results would support avoiding potentially unnecessary invasive procedures.
    • A pragmatic multicentre RCT in a well-defined “at-risk” cohort could clarify whether tracheotomy timing is a modifiable determinant of VAP and clinically important outcomes.

Design & Methods

  • Research Question: Among adult ICU patients ventilated for acute respiratory failure who remain severely impaired at 48 hours without pneumonia, does early (day 6–8) versus late (day 13–15) tracheotomy reduce 28-day cumulative VAP incidence?
  • Study Type: Multicentre, investigator-initiated, open-label, parallel-group randomised controlled trial in 12 ICUs in Italy (June 2004 to June 2008); registered (NCT00262431).
  • Population:
    • Setting: adult ICUs (Italy); mechanically ventilated patients enrolled within 24 hours of starting ventilation.
    • Enrolment criteria (within 24 hours): mechanical ventilation ≥24 hours; no active lung infection (composite pulmonary infection score <6); SAPS II 35–65; SOFA score ≥5.
    • Randomisation criteria (48 hours after enrolment): worsening respiratory condition (PaO2 ≤60 mm Hg on FiO2 0.5 and PEEP 8 cm H2O, or PaO2 ≤100 mm Hg on FiO2 0.9 and PEEP 12 cm H2O); unchanged or worsening SOFA score; no pneumonia.
    • Key exclusions: chronic obstructive pulmonary disease; neck anatomical deformity (including thyromegaly) or cervical tumours; history of oesophageal/tracheal/pulmonary cancer; previous tracheotomy; soft tissue infection of the neck; haematological malignancy; pregnancy.
    • Pre-specified conditions to defer tracheotomy: severe haemodynamic instability; severe arrhythmias; acute myocardial infarction; acute peritonitis; platelet count <50 ×103/µL; bleeding time >2× normal (within 24 hours).
  • Intervention:
    • Early tracheotomy planned for day 6–8 of translaryngeal intubation.
    • Technique/setting not standardised (performed in ICU or operating theatre; percutaneous or surgical at clinician discretion).
    • Co-interventions: semirecumbent positioning protocol; weaning, sedative, and analgesic strategies restricted according to protocol (details not reported in full).
  • Comparison:
    • Late tracheotomy planned for day 13–15 of translaryngeal intubation.
    • Otherwise identical supportive care and prevention bundle elements (including semirecumbency; protocolised approach to sedation/weaning).
    • Many patients in both groups never underwent tracheotomy due to improvement (extubation) or death before the target window.
  • Blinding: Unblinded trial (timing strategy not feasibly masked); for VAP ascertainment, nonobjective composite pulmonary infection score components (radiography and secretion quality) were assessed by independent clinicians blinded to randomisation group.
  • Statistics: A total of 320 patients were required to detect a 35% relative reduction in VAP (from 30% to 20%) with 80% power at the 5% significance level; primary analysis was intention-to-treat, using competing risk methods for 28-day cumulative VAP incidence.
  • Follow-Up Period: 28 days after randomisation for primary/secondary endpoints; hospital outcomes during index admission; one-year survival and long-term care needs reported among hospital survivors with follow-up data available.

Key Results

This trial was not stopped early. Recruitment ran from June 2004 to June 2008 with planned completion.

Outcome Early tracheotomy (n=209) Late tracheotomy (n=210) Effect p value / 95% CI Notes
Ventilator-associated pneumonia by day 28 (primary) 30/209 (14%; 95% CI 10–19) 44/210 (21%; 95% CI 15–26) HR 0.66 95% CI 0.42 to 1.04; P=0.07 28-day cumulative incidence; competing risk analysis
Ventilator-free days at day 28 Median 11 (IQR 0–21) Median 6 (IQR 0–17) Not reported P=0.02 Higher values indicate more days alive and free of mechanical ventilation
ICU-free days at day 28 Median 0 (IQR 0–13) Median 0 (IQR 0–8) Not reported P=0.02 Higher values indicate more days alive and out of ICU
Successful weaning by day 28 161/209 (77%; 95% CI 71–82) 142/210 (68%; 95% CI 61–74) Not reported P=0.002 Defined as disconnection from ventilator for >48 hours
Discharge from ICU by day 28 101/209 (48%; 95% CI 41–55) 82/210 (39%; 95% CI 32–46) Not reported P=0.03 Time-to-event signal also favoured early strategy (see below)
Survival at day 28 154/209 (74%; 95% CI 68–80) 144/210 (68%; 95% CI 62–75) HR 0.80 95% CI 0.56 to 1.15; P=0.25 Hazard ratio refers to death within 28 days
Time to ventilator disconnection within 28 days Not reported Not reported HR 0.70 95% CI 0.56 to 0.87; P not reported Hazard ratio for remaining connected to ventilator (lower = earlier liberation)
Time to ICU discharge within 28 days Not reported Not reported HR 0.73 95% CI 0.55 to 0.97; P not reported Hazard ratio for remaining in ICU (lower = earlier discharge)
Any tracheotomy-related adverse event (among those receiving tracheotomy) 57/145 (39%) 46/119 (39%) Not reported Not reported Most events were minor; individual event comparisons not statistically significant
Hospital length of stay Median 31 days (IQR 17–49) Median 32 days (IQR 18–59) Not reported Not reported Reported as “no difference” without a p value
One-year survival (among patients discharged alive with follow-up) 72/144 (50%; 95% CI 41–61) 63/148 (43%; 95% CI 35–52) Not reported P=0.25 Follow-up data available in 292 hospital survivors
  • VAP incidence was numerically lower with the early strategy (14% vs 21%) but did not meet conventional statistical significance (HR 0.66; 95% CI 0.42 to 1.04; P=0.07).
  • Process outcomes favoured early tracheotomy (ventilator-free days median 11 vs 6; ICU-free days median 0 vs 0 with wider IQR; successful weaning 77% vs 68%).
  • No detectable mortality benefit at 28 days (74% vs 68% survival; P=0.25); tracheotomy-related adverse event rates were 39% in both groups among patients who underwent tracheotomy.

Internal Validity

  • Randomisation and allocation
    • Central computer-generated randomisation with concealed allocation was reported.
    • Randomisation occurred after a 48-hour “enrichment” period, which improves clinical plausibility for prolonged ventilation but introduces a pre-randomisation selection step (600 enrolled; 419 randomised).
  • Attrition, exclusions, and missingness
    • Pre-randomisation exclusions were substantial (181/600), mainly due to improved respiratory status (n=82), moribund/death (n=30), pulmonary infection (n=24), and other reasons (n=45).
    • Post-randomisation “non-receipt” of the assigned procedure was common: 64/209 (31%) in early and 91/210 (43%) in late did not undergo tracheotomy (mostly improvement/extubation or death), diluting any causal effect of the procedure itself.
    • Long-term follow-up outcomes were reported only for hospital survivors with data available (292 patients), which is vulnerable to incomplete capture beyond discharge.
  • Performance and detection bias
    • Open-label design creates risk that clinician behaviour (sedation/weaning decisions, diagnostic work-up for pneumonia) could be influenced by knowledge of group assignment.
    • VAP outcome ascertainment attempted partial blinding: nonobjective composite pulmonary infection score components were assessed by blinded clinicians; nevertheless, VAP diagnosis remains a complex, partly subjective construct.
  • Protocol adherence and separation of the variable of interest
    • Timing separation among those who received tracheotomy was clear: mean 7 (SD 1) days vs 14 (SD 1) days of translaryngeal intubation before tracheotomy.
    • Strategy adherence was incomplete: tracheotomy performed in 145/209 (69%) early vs 119/210 (57%) late.
    • Key co-interventions (semirecumbent positioning; protocol restrictions on sedation/weaning) were specified, aiming to reduce between-group contamination; tracheotomy technique and location were not standardised.
  • Baseline comparability and illness severity
    • Baseline characteristics were well balanced at enrolment: age 61±16 vs 61±15 years; SAPS II 49.8±9.2 vs 49.8±9.5; SOFA 7.8±2.1 vs 7.7±2.0.
    • At randomisation, respiratory severity was similar: PaO2 76±21 vs 73±21 mm Hg; FiO2 0.57±0.16 vs 0.57±0.16; PEEP 10.1±2.2 vs 10.0±2.0 cm H2O.
    • Diagnosis categories were closely matched (e.g., neurological 28% vs 30%; respiratory 22% vs 22%; sepsis 7% vs 7%).
  • Statistical rigour
    • Primary analysis was intention-to-treat, with competing risk methods for VAP (important because death and extubation are competing events for pneumonia ascertainment).
    • Assumed control VAP incidence in the power calculation (30%) exceeded observed incidence in the late group (21%), reducing power for the targeted effect size.

Conclusion on Internal Validity: Overall, internal validity appears moderate: allocation concealment and intention-to-treat analysis support causal inference, but open-label care, substantial non-receipt of tracheotomy, and reliance on a partly subjective VAP construct likely attenuated detectable between-group differences.

External Validity

  • Population representativeness
    • Patients were selected for moderate–high severity early in mechanical ventilation (SAPS II 35–65; SOFA ≥5) and then enriched at 48 hours for persistent/worsening respiratory dysfunction without pneumonia.
    • Key exclusions (notably COPD and significant neck pathology) remove clinically common phenotypes and limit applicability to mixed ICUs with high COPD prevalence.
  • Applicability across settings
    • Conducted in 12 Italian ICUs with protocol elements (semirecumbency; constrained sedation/weaning) that may not be implemented with equal fidelity in all health systems.
    • Because tracheotomy technique and location were not standardised, the “strategy” likely reflects real-world practice variability, improving pragmatic relevance for similar resource settings.
    • Findings are most applicable to patients who remain severely impaired at 48 hours and in whom a prolonged course is anticipated; extrapolation to lower-acuity cohorts is uncertain.

Conclusion on External Validity: Generalisability is moderate: results are most relevant to higher-severity, persistently hypoxaemic ICU patients in systems with established VAP prevention and weaning protocols, but less applicable to COPD-heavy case-mix or settings with different procedural thresholds/resources.

Strengths & Limitations

  • Strengths:
    • Multicentre randomised design with concealed allocation and intention-to-treat analysis.
    • Clinically coherent enrichment approach (48-hour reassessment) focusing on patients plausibly at risk for prolonged ventilation and VAP.
    • Use of competing risk methods for VAP, acknowledging the methodological challenge that death/extubation preclude pneumonia diagnosis.
    • Meaningful patient-centred and system outcomes reported (ventilator-free days, ICU discharge, adverse events; longer-term outcomes among hospital survivors).
  • Limitations:
    • Open-label care with potential for performance bias in sedation, weaning, and diagnostic intensity for VAP.
    • Primary endpoint (VAP) depends on clinical criteria and scoring (composite pulmonary infection score), which may misclassify pneumonia and is susceptible to ascertainment bias despite partial blinding.
    • Substantial non-receipt of tracheotomy in both groups (31% early; 43% late) dilutes the effect of the procedure and shifts inference towards a “timing strategy” rather than a “tracheotomy vs no tracheotomy” effect.
    • Observed control VAP incidence (21%) was lower than assumed (30%), weakening statistical power; the primary endpoint narrowly missed significance (P=0.07).

Interpretation & Why It Matters

  • Infection prevention
    Early tracheotomy did not demonstrate a statistically significant reduction in 28-day VAP (14% vs 21%; HR 0.66; 95% CI 0.42 to 1.04; P=0.07), suggesting that timing alone is unlikely to be a dominant driver of VAP risk in ICUs using standard prevention measures.
  • Liberation and throughput
    Process outcomes favoured earlier timing (ventilator-free days median 11 vs 6; ICU discharge by day 28 48% vs 39), consistent with a strategy that may facilitate earlier liberation/discharge in selected patients—although open-label care and incomplete adherence mean causality could be partly behavioural (weaning/sedation decisions) rather than purely mechanistic.
  • Hard outcomes and harms
    No mortality benefit was detected at 28 days (74% vs 68% survival; P=0.25), and among those who underwent tracheotomy, adverse events occurred in 39% in both groups; therefore, routine acceleration of tracheotomy purely to improve survival is not supported by these data.
  • Clinical decision-making
    The trial supports an individualised approach: early tracheotomy may be reasonable when goals prioritise sedation minimisation, comfort, and a plausible trajectory towards weaning, but the evidence does not justify early tracheotomy as a universal VAP-prevention intervention.

Controversies & Subsequent Evidence

  • Predicting “prolonged ventilation” and the risk of unnecessary procedures
    • The trial’s design highlights prognostic uncertainty: 181/600 enrolled patients were not randomised, and after randomisation 64/209 (31%) early and 91/210 (43%) late never received tracheotomy due to extubation or death before the target window.
    • The accompanying editorial emphasised that imperfect prediction can lead to avoidable tracheotomy exposure when an “early” strategy is adopted broadly.1
  • Endpoint selection and interpretability
    • VAP was chosen as the primary outcome, but its diagnosis is inherently challenging; even with partial blinding of some components, misclassification and ascertainment variation can bias towards null.
    • The study’s observed VAP rate in the control strategy (21%) was lower than the 30% assumed for powering, plausibly contributing to the borderline primary result (P=0.07) and wide confidence interval crossing unity.
  • How later RCT evidence fits
    • The large UK TracMan trial found no survival benefit from early versus late tracheostomy in mechanically ventilated ICU patients, reinforcing the view that routine early timing does not improve hard outcomes.2
  • What systematic reviews conclude
    • Meta-analytic syntheses generally show no consistent mortality benefit of early tracheostomy, with more variable signals for pneumonia incidence and ICU/ventilation duration; heterogeneity in timing definitions and case-mix remains a key limitation.34
  • Guideline positioning
    • Contemporary liberation guidance has evaluated early tracheostomy as a strategy but does not support routine early tracheostomy for all ventilated adults, favouring individualised decision-making alongside protocolised sedation and spontaneous breathing trials.5

Summary

  • Multicentre Italian RCT tested an early (day 6–8) versus late (day 13–15) tracheotomy timing strategy in adult ICU patients with persistent severe respiratory dysfunction at 48 hours and no pneumonia.
  • Primary endpoint (28-day cumulative VAP) was lower with early timing (14% vs 21%) but not statistically significant (HR 0.66; 95% CI 0.42 to 1.04; P=0.07).
  • Early timing improved process outcomes: ventilator-free days median 11 vs 6 (P=0.02); successful weaning 77% vs 68% (P=0.002); ICU discharge by day 28 48% vs 39% (P=0.03).
  • No mortality benefit was detected at 28 days (74% vs 68% survival; P=0.25) or at one year among hospital survivors with follow-up (50% vs 43%; P=0.25).
  • Only 69% (early) and 57% (late) actually underwent tracheotomy; among those receiving tracheotomy, adverse events occurred in 39% in both groups.

Further Reading

Other Trials

Systematic Review & Meta Analysis

Observational Studies

Guidelines

Overall Takeaway

In a selected cohort of mechanically ventilated ICU patients still severely impaired at 48 hours, scheduling tracheotomy at day 6–8 did not significantly reduce VAP compared with day 13–15, but did improve ventilator liberation and ICU discharge metrics. The absence of a mortality signal, the borderline primary endpoint, and substantial non-receipt of tracheotomy reinforce that timing decisions should be individualised rather than protocolised as a universal VAP-prevention strategy.

Overall Summary

  • Early tracheotomy (day 6–8) did not significantly lower VAP (14% vs 21%; P=0.07) but improved ventilator-free days (11 vs 6) and weaning (77% vs 68%), with no mortality benefit and similar procedural adverse event rates.

Bibliography